The main concern I have with ASML is that they're largely a play on Moore's law, which is slowing down, and could come to a halt in the coming decade - you can't scale smaller than the size of an atom. For each new node that TSMC and the likes move to, they need to buy more and more ASML tools, especially EUV, to enable this shrinkage. However, once that growth driver levels off, demand for ASML tools will shrink. The company's tools are actually so good that each one they've sold is still in operation! This is great, but not from a business perspective. I bought a huge chunk of ASML in '16 and sold it in the tech bubble of '21. But definitely a great company.
Aixtron had a similar problem post 2011, they make tools for LED manufacturing. Once the LED fabs were built, demand for their tools plummeted. If/ when Moore's law slows down dramatically, say from 2-3 years to 3-5 years, the demand for ASML tools at the leading node gets spread out over more years, resulting in negative demand growth for ASML's tools. Keeping a healthy pace for Moore's law is key for ASML.
Thanks for an insightful article! Not sure if you got any feedback regarding the the power data?
Anyway, I got curious about it to so did some quick calculations.... Wennink mentioned 24 000 terrawatt hours (TWh) of electricity - which would be the total of energy consumed that year.
To get to the capacity needed to produce that amount, you take the total amount of hours per year (24x365=8760) and then the "efficiency rate" of each power generation technology - wind or solar - and then you can calculate the capacity needed to produce.
So if we use a global average - i.e. wind 26% and solar 14% (average global efficiency data calculated from numbers on capacity and production I found in "Our world in data")
- Assuming that ALL that 24000TWh of energy would be produced from wind it would imply an semis content of roughly EUR 32 bn and if all was solar - EUR 78 bn....
Ya I corrected it a while back, but the email changed. I got feedback from a professional, they think it's more that 70-90B range due to peak vs average power and the correct calculation. Sorry for the confusion. I messed that up bad, but to be fair, i did exclaim i wanted someone to check the work.
Ah, ok. well just wanted to help. And I was curious as well about if renewables really would be a gamechanger for the semis demand going forward. My impression is it does not, even though you put it into that 70-90 bn range perspective. If its 10% of total today, maybe it could reach 40% (?) by 2030 - an increase of 30% - it would not take that big of a chunk of the total semis demand, or am I totally wrong here? I guess ramping up EV production would have a much larger effect, dont you think?
And BTW, whats your take on China's chance to catch up on the semiconductor tech? Even though they are able to make 7nm chips now (less efficiently with DUV?!) and even if they would manage to build some sort of EUV machine sometime in the future... well wouldn't TSMC, Intel and Samsung just be miles ahead already by then?
I am a "newcomer" into semis, so thanks again for that great article!!
And I thank you for it. I practically called out for help because the number didn't make sense to me! Please point out any thoughts you have in the future as well.
I agree with you, although some countries, the transition is much higher/faster.
EVs are bigger because leading edge content, agree.
80% of chips aren't 7nm or lower leading edge, value wise, that a lower figure, but they keep having a larger and larger addressable market.
We still need to address potential competitor to ASML from China. Those sanctions are forcing the Chinese to do something about this. If they don't a whole bunch of industries will collapse. I am sure they don't take that lightly.
They took us by surprise with their 7nm. And now they're developing the 5nm (I think done already)
We don't know exactly where they are in the Lithography development. They're not disclosing anything. It looks like Zeiss is doing some tech transfer there after Scholz Visit to China (a 25 billion deal signed).
They built a space station, went to the moon and Mars. Apparently 30% of ASML owned patent have Chinese names on them. They have as many engineers as we have water here in Canada.
I understand it's mighty complex to build an EUV machine. But they will get it eventually. 5 years max. And then the nightmare will start for ASML and the rest of the players in the industry; because in order to build the EUV machine you will need to build the supply chain associated with it. They will be more efficient and will kill most competition.
And the worst part...we forced them to do it. It was a matter of survival for them. This is not the first time we pushed them into a corner. We're not learning anything. We keep doing the same mistake.
I think politicians in the west should stop smoking weed. It's bad for long term decision making.
What's your guess? 15 years? SMEE builds DUV for quiet some time nowHow long does it take to go from DUV to EUV? What technical expertise is required that the Chinese can't develop or can't hire people for it?
They got the precision laser part done a few years ago and I am not too sure about the optics part.
Since money is no issue for SMEE what is the earliest you think they can get it done?
The main concern I have with ASML is that they're largely a play on Moore's law, which is slowing down, and could come to a halt in the coming decade - you can't scale smaller than the size of an atom. For each new node that TSMC and the likes move to, they need to buy more and more ASML tools, especially EUV, to enable this shrinkage. However, once that growth driver levels off, demand for ASML tools will shrink. The company's tools are actually so good that each one they've sold is still in operation! This is great, but not from a business perspective. I bought a huge chunk of ASML in '16 and sold it in the tech bubble of '21. But definitely a great company.
That makes 0 sense. Demand for chips still grows even if Moore's law stops, requiring more tooling.
Aixtron had a similar problem post 2011, they make tools for LED manufacturing. Once the LED fabs were built, demand for their tools plummeted. If/ when Moore's law slows down dramatically, say from 2-3 years to 3-5 years, the demand for ASML tools at the leading node gets spread out over more years, resulting in negative demand growth for ASML's tools. Keeping a healthy pace for Moore's law is key for ASML.
No it's not comparable. Led market, growth, capital intensity is ridiculously different.
Moore's law slowing down is actually amazing for semicap demand.
Thanks for an insightful article! Not sure if you got any feedback regarding the the power data?
Anyway, I got curious about it to so did some quick calculations.... Wennink mentioned 24 000 terrawatt hours (TWh) of electricity - which would be the total of energy consumed that year.
To get to the capacity needed to produce that amount, you take the total amount of hours per year (24x365=8760) and then the "efficiency rate" of each power generation technology - wind or solar - and then you can calculate the capacity needed to produce.
So if we use a global average - i.e. wind 26% and solar 14% (average global efficiency data calculated from numbers on capacity and production I found in "Our world in data")
- Assuming that ALL that 24000TWh of energy would be produced from wind it would imply an semis content of roughly EUR 32 bn and if all was solar - EUR 78 bn....
Ya I corrected it a while back, but the email changed. I got feedback from a professional, they think it's more that 70-90B range due to peak vs average power and the correct calculation. Sorry for the confusion. I messed that up bad, but to be fair, i did exclaim i wanted someone to check the work.
Ah, ok. well just wanted to help. And I was curious as well about if renewables really would be a gamechanger for the semis demand going forward. My impression is it does not, even though you put it into that 70-90 bn range perspective. If its 10% of total today, maybe it could reach 40% (?) by 2030 - an increase of 30% - it would not take that big of a chunk of the total semis demand, or am I totally wrong here? I guess ramping up EV production would have a much larger effect, dont you think?
And BTW, whats your take on China's chance to catch up on the semiconductor tech? Even though they are able to make 7nm chips now (less efficiently with DUV?!) and even if they would manage to build some sort of EUV machine sometime in the future... well wouldn't TSMC, Intel and Samsung just be miles ahead already by then?
I am a "newcomer" into semis, so thanks again for that great article!!
And I thank you for it. I practically called out for help because the number didn't make sense to me! Please point out any thoughts you have in the future as well.
I agree with you, although some countries, the transition is much higher/faster.
EVs are bigger because leading edge content, agree.
80% of chips aren't 7nm or lower leading edge, value wise, that a lower figure, but they keep having a larger and larger addressable market.
Very detailed.
We still need to address potential competitor to ASML from China. Those sanctions are forcing the Chinese to do something about this. If they don't a whole bunch of industries will collapse. I am sure they don't take that lightly.
Smee is a long ways away
They took us by surprise with their 7nm. And now they're developing the 5nm (I think done already)
We don't know exactly where they are in the Lithography development. They're not disclosing anything. It looks like Zeiss is doing some tech transfer there after Scholz Visit to China (a 25 billion deal signed).
They built a space station, went to the moon and Mars. Apparently 30% of ASML owned patent have Chinese names on them. They have as many engineers as we have water here in Canada.
I understand it's mighty complex to build an EUV machine. But they will get it eventually. 5 years max. And then the nightmare will start for ASML and the rest of the players in the industry; because in order to build the EUV machine you will need to build the supply chain associated with it. They will be more efficient and will kill most competition.
And the worst part...we forced them to do it. It was a matter of survival for them. This is not the first time we pushed them into a corner. We're not learning anything. We keep doing the same mistake.
I think politicians in the west should stop smoking weed. It's bad for long term decision making.
5 years? No shot
What's your guess? 15 years? SMEE builds DUV for quiet some time nowHow long does it take to go from DUV to EUV? What technical expertise is required that the Chinese can't develop or can't hire people for it?
They got the precision laser part done a few years ago and I am not too sure about the optics part.
Since money is no issue for SMEE what is the earliest you think they can get it done?
They don't even have Arf dry working well yet let alone a good Arfi tool with good overlay.
Another masterpiece of analytical reporting. Many thanks.